Will Mark Zuckerberg open the Pandora box?

Meta (formerly Facebook) has announced plans to turn its large language model, LLaMA (Large Language Model Meta AI) into an open source instrument. The decision, according to a NYT report, has at its core the desire to compete with artificial intelligence of other giants such as Google, Microsoft, and especially OpenAI. 

Last February, Meta made the underlying code for LLaMA available for download via email to anyone whom Meta considered trustful, in contrast to its competitors’ strategy to keep such models under lock. 

Meta's chief AI scientist, Yann LeCun, believes that hoarding the code is a dangerous approach.

“Keeping powerful models behind closed doors is a huge mistake. […] Do you want every AI system to be under the control of a couple of powerful American companies?" he emphasized.

Critics argue meanwhile that the challenges of content moderation are already significant, and introducing generative AI adds complexity to the issue. There’s a huge risk for large-scale misuse of AI, which has been caught faking events and spreading disinformation among users.

More to read:
[video] Artificial intelligence created a short film about human stupidity

Nor is Meta a charity organization and whatever the motive behind declassification of the LLaMA code is, Mark Zuckerberg surely doesn’t care about public benefit as much as about Meta’s profits. In 2022 Meta quietly shut down its metaverse project, disappointing investors and users alike. 

Researchers at Stanford University, who obtained the code, ran an experiment with LLaMA to analyze its behaviour. They were shocked of the texts Meta's AI was able to generate. 

Opening up the source code of a potentially-dangerous AI to anyone is like releasing the instructions how to manufacture an atomic bomb to talented but insane scientists.